Founded by two early Datadog engineers and backed by Greylock's Jerry Chen, Niteshift is betting that the real money in AI coding isn't picking the winner. It's running the infrastructure underneath all of them.
When the AI coding market already has Cursor valued at $29 billion and Cognition's Devin agent sitting at $26 billion on the back of a $1 billion raise, you might wonder what the contrarian play looks like. For Sajid Mehmood and Conor Branagan, both early engineers at Datadog who spent nearly a decade building cloud-native infrastructure, the answer is: don't bet on the agent. Bet on the environment the agent runs in. That's the thesis behind Niteshift, which announced a $7 million seed round today led by Greylock partner Jerry Chen, alongside Amplify Partners, BoxGroup, and SV Angel, while simultaneously opening general availability after an earlier invite-only period.
The platform's core proposition is model neutrality by design. Rather than committing to Claude Code or OpenAI's Codex or any of the open-source alternatives gaining traction, Niteshift gives engineering teams a full-stack cloud environment where coding agents of any stripe can run, test, and verify work autonomously against real development infrastructure. When an agent finishes a task, it doesn't just hand back modified files. It returns a pull request with observable evidence that the changes actually work. That distinction matters. Most enterprise teams aren't afraid of AI coding agents in principle; they're afraid of agents that confidently produce code they can't verify.
The historical analogy writes itself: sell picks and shovels during a gold rush. But the better precedent for what Niteshift is attempting may be the container infrastructure wave that preceded today's AI moment. Jerry Chen, now leading this investment, was the Greylock partner who led the firm's Series B into Docker, an early, quiet infrastructure bet that turned out to be far more durable than most of the application-layer companies competing on top of it. That pattern, backing the plumbing while the applications proliferate, has become something of a signature for Chen's thesis at Greylock.
The question is whether infrastructure neutrality holds up as a moat as the market matures. OpenAI and Anthropic are not standing still; both companies have strong incentives to offer managed execution environments for their own agents, which could make third-party infrastructure redundant at the high end of the market. The counterargument Niteshift is making, which tracks with a concern that has been surfacing across enterprise software circles, is that the model makers are also your competitors. As TechCrunch noted in its coverage of the launch, the founders frame this as a bet against Big AI lock-in: companies building products on top of a single AI provider's infrastructure risk finding that provider has launched a competing feature or tightened the relationship terms. A neutral layer that can route between Claude Code, Codex, and whatever ships next is, in this framing, not just a technical convenience but a structural hedge.
Ramp recently published data showing background AI agents are now authoring 30 to 40 percent of the company's pull requests. That kind of adoption at a financially sophisticated company is exactly the signal Niteshift is pointing to. If agent-written code is moving from experiment to default at fast-moving organizations, the pressure on execution environments becomes significant quickly. Who runs the tests? Who manages the credentials? Who ensures the agent isn't touching production in ways no one authorized? These are infrastructure problems, not model problems, and they're the ones Niteshift is built to solve.
What the angel list is actually saying
Beyond the lead investor, the cap table for this round is unusually legible as a signal. Reid Hoffman's participation reflects a long-standing bet on infrastructure layers in developer tooling. More telling are Datadog co-founders Olivier Pomel and Alexis Lê-Quôc, who are essentially endorsing their former engineers' ability to build the kind of durable, operationally serious platform Datadog itself became. Ankur Goyal from Braintrust, which focuses on AI evaluation infrastructure, and Misha Laskin from Reflection AI round out a group that collectively has very specific views on where the reliability gaps in AI-generated software currently live. This isn't a group of generalist tech angels taking small positions on a hot sector; it reads more like a group of practitioners who looked at the problem and recognized the gap Niteshift is targeting.
The Datadog pedigree itself deserves attention as a VC signal in its own right. Datadog's origin story is one of the more studied examples of what happens when cloud-native infrastructure specialists time a market transition correctly. The company went from a monitoring tool to a multi-billion-dollar observability platform by building something that sat underneath other people's software stacks and became essential as those stacks grew more complex. Mehmood and Branagan lived that scaling arc from the inside, which gives them credibility not just in infrastructure design but in the go-to-market motion of selling to engineering organizations that need reliability before they need features.
The AI coding market is moving from proof-of-concept to production, and the shift is happening faster than most infrastructure caught up with the last wave. Niteshift's timing puts it at the beginning of that gap widening, not after it closes. The company to watch isn't necessarily the one writing the most code. It's the one that makes any agent's code trustworthy enough to actually ship.
Also read: Jedify raises $24M to build the context layer that gives enterprise AI agents something to actually work with • German humanoid startup Neura Robotics closes the largest robotics funding round in history as Tether redeploys its stablecoin war chest into physical AI • China's factory inflation puts new pressure on the AI buildout